Biometrical Letters Vol. 44(2), 2007, pp. 105-128
|
ROBUSTNESS OF JOINT REGRESSION ANALYSIS
D.G. Pereira1, J.T. Mexia2, P.C. Rodrigues2 1Department of Mathematics, University of Évora, CIMA-UE (Center for Research on Mathematics and its Applications), Colégio Luís António Verney, Rua Rom?o Ramalho, 59, 7000-671 Évora, Portugal, e-mail: dgsp@uevora.pt 2Department of Mathematics, Faculty of Sciences and Technology, Nova University of Lisbon, CMA-UNL (Center for Mathematics and its Applications), Quinta da Torre, 2825-516 Monte da Caparica, Portugal |
Joint Regression Analysis is shown to be extremely robust to missing observations. Thus, using a series of "alfa-designs" of winter rye cultivars, it was shown that with up to 40% of missing observations the cultivars selected would be the same. In this study we considered missing observations incidences varying from 5% to 75%, with a step size of 5%. For each incidence the positions of missing observations were randomly generated in triplicate.
Joint Regression Analysis; Robustness; missing observations; Linear regressions; L2 environmental indexes.
